World Applied Sciences Journal 20 (7): 946-954, 2012 ISSN 1818-4952 © IDOSI Publications, 2012 DOI: 10.5829/idosi.wasj.2012.20.07.1771 Corresponding Author: Khairul Salleh Mohamed Sahari, College of Engineering, Universiti Tenaga Nasional (UNITEN), 43000 Kajang, Selangor, Malaysia. Tel: +60-389212020. 946 Comparison Between Genetic Algorithm and Electromagnetism-Like Algorithm for Solving Inverse Kinematics Issa Ahmed Abed, S.P. Koh, Khairul Salleh Mohamed Sahari, 1 1 1 S.K. Tiong and David F.W. Yap 1 2 College of Engineering, Universiti Tenaga Nasional, Selangor, Malaysia 1 Faculty of Electronic and Computer Engineering, 2 Universiti Teknikal Malaysia Melaka, Melaka, Malaysia Abstract: A comparison study between Electromagnetism-Like Algorithm (EM) and Genetic Algorithm (GA) has been presented in this work to solve the Inverse Kinematics (IK) of a four-link planar robot manipulator. The comparison is focused on some points for both algorithms like the accuracy of the results and the speed of convergence. Different target points have been taken to check the performance of each algorithm to solve the IK problem. The results showed that EM algorithm needs less population size and number of generations to get the true solution. There are multiple robot configurations at the goal points and both algorithms are able to find these solutions at each point. Self developed software simulator is used to display some of these solutions at each goal position. Key words: Inverse kinematics Real coded genetic algorithm Attraction-repulsion mechanism Planar manipulator INTRODUCTION kinematics for a three joint robot. The initial and final Nowadays, there are many applications for robots. After that, all the angles that obtained from (x, y, z) These robots have non-linear kinematics equations. coordinates are recorded in a file named as training set of Their inverse kinematics solution provides the joint neural network. Based on metaheuristics algorithms, angles which are required to attain a particular position of Chandra and Rolland [5] proposed hybrid algorithm based the robot wrist in the robot work space [1]. The mapping on genetic algorithm and Simulated Annealing (SA) to from joint space to the end effector space is referred to as solve the forward kinematics of the 3RPR parallel Forward Kinematics (FK). Finding the joint angle of the manipulator. In this method, both algorithms are manipulator from end effector position is referred as hybridized into two hybrid metaheuristic techniques. Inverse Kinematics (IK). The forward kinematics One of the limitations in this method is long optimization equations can be solved easily, but it is difficult to solve time. Two examples for SCARA and PUMA robots have inverse kinematics exactly for high-order degree of been taken by Kalra et al. to check an evolutionary freedom. approach based on real-coded genetic algorithm to get the Many approaches have been proposed to solve solution of the multimodal inverse kinematics problem [1]. inverse kinematics equations. One of these approaches is In the last years, new metaheuristic methods such as to use numerical methods [2]. In numerical methods, EM algorithm have been used to solve the inverse a good initial guess must be given because these methods kinematics for robot manipulators. EM algorithm is an are divergence and vulnerable to local optimums [3]. optimization algorithm that uses the principle of Recently, artificial intelligence methods were applied attraction-repulsion mechanism. EM algorithm has to solve the inverse kinematics problem. Kõker et al. [4] been applied to different problems, such as have designed a neural network to solve the inverse optimizations problems [6], scheduling problems [7] and points have been generated by using cubic polynomial.